1. Field of the Invention
This invention relates to replaying writes, and more particularly, to replaying writes in a replication log.
2. Description of the Related Art
A replication system typically sends application writes performed on a primary volume to a secondary volume for replication. This replication is performed in order to maintain a consistent copy of application data at the secondary volume.
Some types of replication involve application writes being queued in a replication log before an application write is sent to the secondary volume for replication. Problems may arise when the number of application writes queued in a replication log increases to an amount close to or beyond the capacity of a replication log. When this occurs, the application data within the secondary volume may become unsynchronized with the primary volume. A replay of the application writes within the replication log is then performed to resynchronize the application data between the primary and secondary volume. During a replay, the regions of the application data that differ between a primary volume and secondary volume are identified and resynchronized.
Typically, this replay process is lengthy and may consume large amounts of computing resources. In addition, the replay process may also delay the completion of subsequent application writes to a primary volume.